Solutions · Agency
One console for the whole client roster
Built for teams running experimentation as a service. Every account in one place, ordered by what needs a person first, with reporting your client can read.
Across the roster
The morning routine, the report, the paperwork
Every client account on one screen, results a client can read without you translating them, and traffic pooled across all of them.
The roster
Ninety websites, one morning routine
The client list ranks accounts by problems first, then by decisions waiting. You open one screen and know where the day goes.
- Accounts ordered by what needs you
- Scope the whole console to one client in two clicks
- Per-account running counts and win rate
Reporting
A result the client can read without you in the room
Every figure in the console carries its own explanation, written to say what the number means rather than to define the term, and the verdict is stated in words as well as in maths. Half of a retainer review is usually translation. This is the half you get back.
- A plain-English explanation on every statistic, in place
- The verdict in words, next to the p-value that produced it
- Read-only logins, so a client can watch without being able to change anything
- Decided experiments export with the winner, the goal, the lift and the p-value
- Revenue and average order value in the client currency
Commercials
Pooled traffic and paperwork that scales
Traffic is pooled across the roster rather than metered per client, and the agreements are written for an agency: one DPA, your own subprocessor position, SSO for the team.
- Pooled tested traffic across every client account
- Custom DPA and SSO on the agency plan
- Unlimited client accounts
Across the roster
What ninety websites need that one website does not
A single account is a screen you read. A roster is a search problem, an access problem and a coordination problem, and those are the three the console is shaped around.
One rollup, every client
What is broken, what needs a decision and what is queued, grouped by client and by website, across every account you can reach. It is the shape of a Monday morning rather than a dashboard you have to interpret.
Scope comes from the session
An agency sees its client roster; a direct customer sees their own account. There is no client identifier in the address to check, and therefore none to tamper with.
No date picker, on purpose
The roster view reports on what is running now, and none of its signals are windowed. A broken split is broken, a test past its end date is late, a dead one is dead. Volume over a period is a different screen.
Collisions are checked per site
Two of your consultants working on the same client can put two variants on the same element without either of them knowing. The pair is reported before it can be automated, on the same rules the scheduler enforces.
Concurrency scales with the roster
Because experiments randomise independently, a client site can carry as many tests as it has ideas without any of them concluding more slowly. Throughput per account is what makes a roster profitable.
Everyone reads their own layout
Console layouts are saved per person and per screen, so an agency lead scanning for problems and an analyst reading one experiment do not have to share a view or rebuild it every morning.
Questions
What agencies ask first
How many clients can one consultant realistically carry?
More than the tooling usually allows, for two reasons worth checking against what you run today. Concurrency does not divide traffic here, so a client site can carry as many live tests as it has ideas rather than queueing them one at a time, which is the single biggest constraint on what a retainer can deliver. And you do not audit the roster to find problems: everything unresolved across every account is rolled up in one view, ranked so that anything untrustworthy comes before anything merely waiting on a decision. The morning check is a screen rather than a circuit of accounts.
Can consultants be limited to their own clients?
Access is resolved from the signed-in user and their company, so an agency sees its own roster and nothing else on the platform. Within that, roles decide what somebody can do rather than what they can see: members are read-only everywhere, admins and owners can start, stop and roll out. Every change is recorded against the experiment with a name and a timestamp, which is what makes a shared account across a team auditable rather than merely convenient.
What can we actually put in front of a client?
Two things, and the first is the one agencies underrate. Every figure in the console explains itself in plain English, written to say what the number means rather than to define the term, and the verdict is stated in words next to the maths that produced it. A client with a read-only login can therefore follow a result without you translating it, and read-only means exactly that: they can see everything and change nothing. The second is the export, which carries every decided experiment with the winning variant, the goal it won on, the lift and the p-value, which is what their developer needs. Branded reports with your own logo and colours are not in the product today, so if a designed monthly deck is central to how you sell, plan on building it from the export.
What about clients whose traffic is modest?
This is the conversation worth having before the retainer rather than three months into it, and the platform is unusually direct about it. Every experiment reports the smallest difference its traffic can reliably detect, so a site that cannot answer the question you are asking says so instead of sitting at "not significant" for two months while you defend the invoice. On a small client that changes what you sell them: bolder tests, higher-traffic templates, or a different service.
Is it safe to let it run experiments without us watching?
It defaults to not doing so, which is the answer you want on somebody else’s site. Automation and automatic rollout of a winner are separate switches, both off until you turn them on, and both refuse to act on data they cannot trust. With automation off the engine still evaluates every board and records what it would have done, using the same code that would have done it, so you can watch it be right across a client roster for a month before you let it touch anything.